[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: g95 compilation of netcdf

Matthieu Verstraete <address@hidden> writes:

> Hi Ed,
>
> on a brand new install of FC5 with the latest g95 (neither of which
> should influence what I'm about to say) netcdf finds FFLAGS="" and
> correctly compiles the whole way through. Yay. There must have been
> some crap left in my build environment on the other machine - another
> argument in favor of making changes to configure only on the command
> line :). It does make fortransymbol__, however, so I've re-made it for
> my own purposes.
>
> Ciao
>
> Matthieu

OK, great. Thanks so much for letting me know.

>
> PS: what's the status on netcdf handling files/variables bigger than 2
> or 4Gb? I know with the present version only the last variable can be
> bigger, and I need to have all the large file stuff etc... But there

Starting in version 3.6.0, we introduced 64-bit offset format, an
alternative netCDF format in which you can create very large
files. For more info, including the new size limits of 64-bit offset
format, see:

http://www.unidata.ucar.edu/software/netcdf/docs/netcdf/Large-File-Support.html

Since you just installed 3.6.1, you are good to go. When you create a
file, use the NC_64BIT_OFFSET flag in the cmode. That will create a
file which can handle much larger variable and file sizes. (Note that
it will also be unreadable to versions of netCDF previous to 3.6.0.)

> was talk of extending the limits, and allowing any number of variables
> larger sizes. One colleague went through your sources and found a
> (apparently arbitrarily set) variable for the number of variables, and
> their maximum size etc... We definitely didn't want to start
> distributing our own home binaries for the libs - we'd rather use the
> official distribution. Is there someone in charge of the development
> planning and user needs etc? You, perhaps? Thanks again.

The number of variables in netcdf.h is currently set to 8192. This
number has been arbitrarily increased with almost every release. It is
really intended as more of a convenience for third-party applications
which use the netCDF library.

How many variable would you like to create at once? 

I think the library code itself should be OK with very large numbers
of variables (up to 2**31 of them), but I would want to test this
*very carefully* before using it to store anyone's scientific data.

If you can give me a number, I would have a better idea of what is
possible.

Thanks!

Ed
-- 
Ed Hartnett  -- address@hidden